Jonathan M. Holmes MD
Spectacles, patching and atropine have strong evidence as effective treatments for amblyopia, whereas many other treatments must be regarded with a degree of scepticism until they have been studied in rigorous randomised controlled trials, according to Jonathan M. Holmes MD, Professor and Chair of Ophthalmology and Vision Science, University of Arizona-Tucson, USA.
“Clearly, initial pilot studies are very useful to show potential proof of concept, but we need randomised clinical trials so that we don’t all jump on the latest bandwagon and start treating amblyopia without appropriate evidence,” Prof Holmes told the told the 2020 WSPOS Virtual Meeting.
He noted that in many small case series, amblyopic patients appear to improve in response to treatment, but closer scrutiny shows the response was in fact illusory because of poor study design. Often the studies lack controls or are not randomised and suffer from other design flaws.
In uncontrolled case series, observer bias can influence results, simply because the clinicians naturally want their patients to improve and will be inclined to look for evidence confirming that hope. In addition, in many case series, data are influenced by regression to the mean, resulting from test-retest variability. A patient’s visual acuity will vary somewhat from day to day, and they tend to be enrolled into a study when the visual acuity is worse than a specific threshold, i.e. on day when the vision is poor. On a subsequent day their visual acuity may appear to have improved, when, in reality, it is just an artefact of test-retest variability.
Furthermore, a study’s design may fail to take into account such factors as the natural history of the disease; some patients will improve without treatment. Study data may also be influenced by other treatments patients may be simultaneously receiving, he pointed out.
“To avoid the problems of apparent improvement due to natural history and regression to the mean we should have simultaneous controls. To avoid observer bias we should have masked examiners and to avoid the problems of other treatments we should have a defined protocol,” Prof Holmes said.
However, even with controls, allocation bias – consciously or unconsciously choosing patients for the control or treatment group – must also be avoided, he stressed. That can be achieved with randomisation, which will result in known and unknown confounding factors being equally distributed between treatment and control groups.
“We need to move rapidly from a case series to a randomised clinical trial, with masking of outcome assessment, before we can conclude definitively that a new treatment is effective,” he added.
Trial duration can also influence the interpretation of study results, Prof Holmes said. As an example, he cited a randomised study conducted by the Paediatric Investigator Group (PEDIG) that compared occlusion for at least six hours a day with an atropine eye drop each day in a cohort of amblyopic patients.
The six-month assessment showed similar mean visual acuity gains of 3.16 logMAR lines in the patching group and 2.84 logMAR in the atropine group, concluding that the treatment effect was similar. However, if the trial had only lasted five weeks it would have shown that the patching group had a significantly greater improvement; a mean improvement of 2.22 logMAR lines in the patching group, compared with a mean improvement of 1.37 logMAR lines in the atropine group.
“If we had stopped the trial at five weeks, we would have had an entirely different conclusion, that patching was superior to atropine,” he added.
In order to minimise the potential bias of study duration it is important to have some pilot data on the time course of treatment effect and advocates for each treatment should be involved during the study planning process.
There is strong evidence based on well-designed studies that spectacles alone are a profoundly effective treatment for strabismic, anisometropic and combined amblyopia, Prof Holmes said. He cited a randomised controlled trial, led by Michael Clarke in the UK, comparing no treatment to spectacles alone in amblyopes, with either straight eyes or a microtropia, aged three-to-five years with a visual acuity of 6/9 to 6/36 in their worst eye and 6/6 in their better eye. Masked assessment of best-corrected VA at 54 weeks found modest improvement with no treatment, but mean improvement with spectacles alone was 0.85 logMAR lines greater than those with no treatment. (
Clarke et al BMJ 2003;327;1251).
However, “spectacles alone” result in resolution of amblyopia in only around a quarter to a third of patients, and in patients who do not fully respond, part-time patching should be considered as the next step with the strongest evidence, he said. He cited a randomised PEDIG study that compared two hours a day of prescribed patching to patching prescribed six hours a day in children three-to-seven years of age. Masked assessment at four months showed that visual acuity in both groups had improved from a baseline mean value of 20/63 to 20/32. (
PEDIG, Arch Opthalmol 2003 ;121:603).
Prof Holmes noted that, based on these studies, he now starts his patients with anisometropic, strabismic or combined amblyopia, on spectacles alone. If that does not result in resolution of the condition, he then moves patients on to low-dose patching (two hours/day) or weekend atropine (1%) drops.
